Bad things happen in distributed data processing and if we're prepared for them, it's better. To prevent against such issues Apache Spark is able to recompute failed partition but also to store the computation snapshot as a checkpoint. Both properties apply to GraphX module's fault-tolerance mechanism.
One of previous posts talked about checkpoint types in Spark Streaming. This one focuses more on one type of them - metadata checkpoint.
Checkpoint allows Spark to truncate dependencies on previously computed RDDs. In the case of streams processing their role is extended. In additional, they're not a single method to prevent against failures.
